A series of experiments to discover the optimal learning situation were conducted using computer-assisted instruction based on a mathematical model. The instructional systems were adaptive in that the sequence of instruction varied according to a given student's performance history, and the program was organized to modify itself automatically as more students completed their course and their records were fed back. Courses taught included computer programing for the junior-college level, reading for the first three grades of elementary school, and a foreign-language vocabulary program for colleges. It was concluded that an optimal learning situation could be designed if the following elements were clearly specified: (1) the set of admissible instruction actions; (2) the instructional objectives; (3) a measurement scale that permits costs to be assigned to each of the instructional actions and payoffs to the achievements of instructional objectives; (4) a model of the learning process. It was further concluded that a learning situation will result only if it is realized that the instructional theory must be modified to fit the learning process and that the theory of instruction will influence research on learning. (SK)